1 Setup

Libraries

suppressPackageStartupMessages({
  library(data.table)
  library(DESeq2)
  library(gplots)
  library(here)
  library(hyperSpec)
  library(parallel)
  library(pander)
  library(plotly)
  library(tidyverse)
  library(tximport)
  library(vsn)
  library(zinbwave)
  library(Rtsne)
})

Helper functions

source(here("UPSCb-common/src/R/featureSelection.R"))

Graphics

hpal <- colorRampPalette(c("blue","white","red"))(100)

Metadata Sample information

samples <- read_csv(here("doc/samples_final.csv"))

## 
## ── Column specification ────────────────────────────────────────────────────────
## cols(
##   ScilifeID = col_character(),
##   SubmittedID = col_character(),
##   Stages = col_character(),
##   Description = col_character(),
##   ID = col_character()
## )

2 Raw data

filelist <- list.files(here("data/Salmon"), 
                          recursive = TRUE, 
                          pattern = "quant.sf",
                          full.names = TRUE)

Sanity check to ensure that the data is sorted according to the sample info

filelist <- filelist[match(samples$ScilifeID,sub("_sortmerna.*","",basename(dirname(filelist))))]

stopifnot(all(match(sub("_sortmerna.*","",basename(dirname(filelist))),
                    samples$ScilifeID) == 1:nrow(samples)))

name the file list vector

names(filelist) <- samples$ID

Read the expression at the gene level

counts <- suppressMessages(round(tximport(files = filelist, 
                                  type = "salmon",
                                  txOut=TRUE)$counts))

combine technical replicates

samples$ID <- sub("_L00[1,2]", "",
                  samples$ScilifeID)
counts <- do.call(
  cbind,
  lapply(split.data.frame(t(counts),
                          samples$ID),
         colSums))

csamples <- samples[,-1]
csamples <- csamples[match(colnames(counts),csamples$ID),]

read the expression for the pool of lincRNAs we found

linc_read <- read_delim("~/Git/lncRNAs/doc/time_expression_nc_filtered.tsv",
                        delim = " ")

## 
## ── Column specification ────────────────────────────────────────────────────────
## cols(
##   Transcript.ID = col_character(),
##   score = col_double(),
##   S1 = col_double(),
##   S2 = col_double(),
##   S3 = col_double(),
##   S4 = col_double(),
##   S5 = col_double(),
##   S6 = col_double(),
##   S7 = col_double(),
##   S8 = col_double(),
##   maxn = col_double(),
##   n = col_double(),
##   peak = col_character()
## )

linc <- linc_read$Transcript.ID

counts <- counts[linc, ]

2.1 Quality Control

Check how many genes are never expressed

sel <- rowSums(counts) == 0
sprintf("%s%% percent (%s) of %s genes are not expressed",
        round(sum(sel) * 100/ nrow(counts),digits=1),
        sum(sel),
        nrow(counts))

## [1] "0% percent (0) of 169731 genes are not expressed"

Let us take a look at the sequencing depth, colouring by Stages

dat <- tibble(x=colnames(counts),y=colSums(counts)) %>% 
  bind_cols(csamples)

ggplot(dat,aes(x,y,fill=csamples$Stages)) + geom_col() + 
  scale_y_continuous(name="reads") +
  theme(axis.text.x=element_text(angle=90,size=4),axis.title.x=element_blank())

Display the per-gene mean expression

i.e. the mean raw count of every gene across samples is calculated and displayed on a log10 scale.

The cumulative gene coverage is as expected, considering we have lincRNAs, caracterised by a really low signal.

ggplot(data.frame(value=log10(rowMeans(counts))),aes(x=value)) + 
  geom_density() + ggtitle("gene mean raw counts distribution") +
  scale_x_continuous(name="mean raw counts (log10)")

The same is done for the individual samples colored by Stages.

dat <- as.data.frame(log10(counts)) %>% utils::stack() %>% 
  mutate(Stages=csamples$Stages[match(ind,csamples$ID)])

ggplot(dat,aes(x=values,group=ind,col=Stages)) + 
  geom_density() + ggtitle("sample raw counts distribution") +
  scale_x_continuous(name="per gene raw counts (log10)")

## Warning: Removed 3230200 rows containing non-finite values (stat_density).

2.2 Export

dir.create(here("data/analysis/salmon"),showWarnings=FALSE,recursive=TRUE)
write.csv(counts,file=here("data/analysis/salmon/raw-unormalised-gene-expression_data_linc.csv"))

3 Data normalisation

3.1 Preparation

For visualization, the data is submitted to a variance stabilization transformation using DESeq2. The dispersion is estimated independently of the sample tissue and replicate.

csamples$Stages <- factor(csamples$Stages)

there are a lot of zeros, so we use zinbwave

se <- SummarizedExperiment(assays=list(counts=as.matrix(counts)),
                           colData=as.data.frame(csamples))

zinb <- zinbwave(se,K=0,epsilon=1e12,
                 X="~Stages",
                 observationalWeights=TRUE)

save(zinb,file=here("data/analysis/salmon/zinb.rda"))

Check the size factors (i.e. the sequencing library size effect)

dds <- DESeqDataSet(zinb,design=~Stages)

## converting counts to integer mode

dds <- DESeq(dds, 
             sfType = "poscounts", 
             useT = TRUE, 
             minmu = 1e-6)

## estimating size factors

## estimating dispersions

## gene-wise dispersion estimates

## Warning in getAndCheckWeights(object, modelMatrix, weightThreshold = weightThreshold): for 12918 row(s), the weights as supplied won't allow parameter estimation, producing a
##   degenerate design matrix. These rows have been flagged in mcols(dds)$weightsFail
##   and treated as if the row contained all zeros (mcols(dds)$allZero set to TRUE).
##   If you are blocking for donors/organisms, consider design = ~0+donor+condition.

## mean-dispersion relationship

## final dispersion estimates

## fitting model and testing

save(dds,file=here("data/analysis/salmon/dds_linc.rda"))

3.2 Variance Stabilising Transformation

vsd <- varianceStabilizingTransformation(dds, blind=TRUE)
vst <- assay(vsd)
vst <- vst - min(vst)
save(vst,file=here("data/analysis/DE/vst-blind_linc.rda"))

3.3 Variance Stabilising Transformation

vsda <- varianceStabilizingTransformation(dds, blind=FALSE)
vsta <- assay(vsda)
vsta <- vsta - min(vsta)
save(vsta,file=here("data/analysis/DE/vst-aware_linc.rda"))

# prepare the data to build the network
#ID <- rownames(vsta)
#vsta <- cbind(ID,vsta)
#vsta_tibble <- as_tibble(vsta)
#write_tsv(vsta_tibble,path=here("data/analysis/DE/vst-aware_linc.tsv"))
#thing <- read_tsv(here("data/analysis/DE/vst-aware_linc.tsv"))
#library(Biostrings)
#seq <- readDNAStringSet("data/trinity/Trinity.fasta")
#names(seq) <- sub(" .*","",names(seq))
#IDs <- thing$ID
#writeXStringSet(seq[IDs],file="~/Git/lncRNAs/data/analysis/DE/linc_all.fasta")

Validation Check the variance stabilisation. It could be worse, considering we have a pool of lincRNAs.

meanSdPlot(log2(counts(dds)[!mcols(dds)$allZero,]+1))

meanSdPlot(log2(assay(zinb)+1))

meanSdPlot(vst[rowSums(vst)>0,])

meanSdPlot(vsta[rowSums(vsta)>0,])

3.4 QC on the normalised data

3.4.1 PCA

pc <- prcomp(t(vsta))
percent <- round(summary(pc)$importance[2,]*100)

Cumulative components effect

We define the number of variable of the model

nvar=1

And the number of possible combinations

nlevel=nlevels(dds$Stages)

We plot the percentage explained by the different components, the red line represent the number of variable in the model, the orange line the number of variable combinations.

ggplot(tibble(x=1:length(percent),y=cumsum(percent)),aes(x=x,y=y)) +
  geom_line() + scale_y_continuous("variance explained (%)",limits=c(0,100)) +
  scale_x_continuous("Principal component") + 
  geom_vline(xintercept=nvar,colour="red",linetype="dashed",size=0.5) + 
  geom_hline(yintercept=cumsum(percent)[nvar],colour="red",linetype="dashed",size=0.5) +
  geom_vline(xintercept=nlevel,colour="orange",linetype="dashed",size=0.5) + 
  geom_hline(yintercept=cumsum(percent)[nlevel],colour="orange",linetype="dashed",size=0.5)

## Warning: Removed 2 row(s) containing missing values (geom_path).

3.4.2 2D

pc.dat <- bind_cols(PC1=pc$x[,1],
                    PC2=pc$x[,2],
                    csamples)

p <- ggplot(pc.dat,aes(x=PC2,y=PC1,col=Stages,text=dds$ID)) + 
  geom_point(size=2) + 
  ggtitle("Principal Component Analysis",subtitle="variance stabilized counts")

ggplotly(p) %>% 
  layout(xaxis=list(title=paste("PC2 (",percent[2],"%)",sep="")),
         yaxis=list(title=paste("PC1 (",percent[1],"%)",sep="")))

3.4.3 Heatmap

Filter for noise

conds <- factor(csamples$Stages)
sels <- rangeFeatureSelect(counts=vsta,
                           conditions=conds,
                           nrep=3)

vst.cutoff <- 1

Heatmap of “all” genes

mar <- par("mar")

par(mar=c(0.05,0.05,0.05,0.05)) 
hm <- heatmap.2(t(scale(t(vsta[sels[[vst.cutoff+1]],]))),
                distfun=pearson.dist,
                hclustfun=function(X){hclust(X,method="ward.D2")},
                labRow = NA,trace = "none",
                labCol = conds,
                col=hpal)

plot(as.hclust(hm$colDendrogram),xlab="",sub="",labels=conds)

3.5 Conclusion

# The Biological QA is good, considering it's based on lincRNAs. We have no outliers. 
# The sequencing depth decreased comparing to the previous analysis.
# Looking at the PCA, it could be interesting to do DE analysis to see in particular what's going on
# in S3 and S6. I consider those two stages relevant, because I think things are changes here. 
#

4 Session Info

## R version 4.0.3 (2020-10-10)
## Platform: x86_64-pc-linux-gnu (64-bit)
## Running under: Ubuntu 18.04.5 LTS
## 
## Matrix products: default
## BLAS/LAPACK: /usr/lib/x86_64-linux-gnu/libopenblasp-r0.2.20.so
## 
## locale:
##  [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
##  [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8    
##  [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8   
##  [7] LC_PAPER=en_US.UTF-8       LC_NAME=C                 
##  [9] LC_ADDRESS=C               LC_TELEPHONE=C            
## [11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       
## 
## attached base packages:
##  [1] grid      parallel  stats4    stats     graphics  grDevices utils    
##  [8] datasets  methods   base     
## 
## other attached packages:
##  [1] Rtsne_0.15                  zinbwave_1.12.0            
##  [3] SingleCellExperiment_1.12.0 vsn_3.58.0                 
##  [5] tximport_1.18.0             forcats_0.5.1              
##  [7] stringr_1.4.0               dplyr_1.0.4                
##  [9] purrr_0.3.4                 readr_1.4.0                
## [11] tidyr_1.1.2                 tibble_3.0.6               
## [13] tidyverse_1.3.0             plotly_4.9.3               
## [15] pander_0.6.3                hyperSpec_0.99-20201127    
## [17] xml2_1.3.2                  ggplot2_3.3.3              
## [19] lattice_0.20-41             here_1.0.1                 
## [21] gplots_3.1.1                DESeq2_1.30.0              
## [23] SummarizedExperiment_1.20.0 Biobase_2.50.0             
## [25] MatrixGenerics_1.2.1        matrixStats_0.58.0         
## [27] GenomicRanges_1.42.0        GenomeInfoDb_1.26.2        
## [29] IRanges_2.24.1              S4Vectors_0.28.1           
## [31] BiocGenerics_0.36.0         data.table_1.13.6          
## 
## loaded via a namespace (and not attached):
##  [1] colorspace_2.0-0       ellipsis_0.3.1         rprojroot_2.0.2       
##  [4] XVector_0.30.0         fs_1.5.0               rstudioapi_0.13       
##  [7] hexbin_1.28.2          farver_2.0.3           affyio_1.60.0         
## [10] bit64_4.0.5            AnnotationDbi_1.52.0   lubridate_1.7.9.2     
## [13] splines_4.0.3          cachem_1.0.3           geneplotter_1.68.0    
## [16] knitr_1.31             jsonlite_1.7.2         broom_0.7.4           
## [19] annotate_1.68.0        dbplyr_2.1.0           png_0.1-7             
## [22] BiocManager_1.30.10    compiler_4.0.3         httr_1.4.2            
## [25] backports_1.2.1        assertthat_0.2.1       Matrix_1.3-2          
## [28] fastmap_1.1.0          lazyeval_0.2.2         limma_3.46.0          
## [31] cli_2.3.0              htmltools_0.5.1.1      tools_4.0.3           
## [34] affy_1.68.0            gtable_0.3.0           glue_1.4.2            
## [37] GenomeInfoDbData_1.2.4 Rcpp_1.0.6             softImpute_1.4        
## [40] cellranger_1.1.0       vctrs_0.3.6            preprocessCore_1.52.1 
## [43] crosstalk_1.1.1        xfun_0.21              testthat_3.0.1        
## [46] rvest_0.3.6            lifecycle_0.2.0        gtools_3.8.2          
## [49] XML_3.99-0.5           edgeR_3.32.1           zlibbioc_1.36.0       
## [52] scales_1.1.1           hms_1.0.0              RColorBrewer_1.1-2    
## [55] yaml_2.2.1             memoise_2.0.0          latticeExtra_0.6-29   
## [58] stringi_1.5.3          RSQLite_2.2.3          highr_0.8             
## [61] genefilter_1.72.1      caTools_1.18.1         BiocParallel_1.24.1   
## [64] rlang_0.4.10           pkgconfig_2.0.3        bitops_1.0-6          
## [67] evaluate_0.14          labeling_0.4.2         htmlwidgets_1.5.3     
## [70] bit_4.0.4              tidyselect_1.1.0       magrittr_2.0.1        
## [73] R6_2.5.0               generics_0.1.0         DelayedArray_0.16.1   
## [76] DBI_1.1.1              pillar_1.4.7           haven_2.3.1           
## [79] withr_2.4.1            survival_3.2-7         RCurl_1.98-1.2        
## [82] modelr_0.1.8           crayon_1.4.1           KernSmooth_2.23-18    
## [85] rmarkdown_2.6          jpeg_0.1-8.1           locfit_1.5-9.4        
## [88] readxl_1.3.1           blob_1.2.1             reprex_1.0.0          
## [91] digest_0.6.27          xtable_1.8-4           munsell_0.5.0         
## [94] viridisLite_0.3.0

LincRNAs Biological QA

Camilla Canovi

2021-02-11